Risk model for central nervous system involvement in patients with diffuse large B cell lymphoma: Temporal validation study

Diffuse large B-cell lymphoma (DLBCL) is the most common lymphoid malignancy in adulthood, and central nervous system (CNS) involvement develops in approximately 5% of patients during follow-up. Accurate identification of risk groups for CNS recurrence is critical for screening and prophylaxis approaches. The aim of the study is to develop a temporally validated risk model predictive of CNS recurrence in DLBCL patients. 460 patients followed in our center between January 2001 and December 2014 were defined as the "derivation" cohort, and 204 patients followed between January 2015 and January 2023 were defined as the "validation" cohort. International prognostic index and extranodal involvements were examined retrospectively. Model parameters were determined by Cox regression analysis and the risk score was calculated based on the Framingham Heart Study scoring system. As a result, a risk model consisting of ≥2 extranodal site involvement, poor performance score and at least 1 risky extranodal site involvement was developed (Omnibus test: -2 log-odds=256.1; χ2=91.2; p<0.001). The developed risk model significantly predicted CNS recurrence (AUC: 0.852; p<0.001). CNS recurrence rates were found to be 1.5% in the low risk group, 6.2% in the medium risk group, and 38.9% in the high risk group. The sensitivity and specificity of the high-risk category in detecting CNS recurrence were determined as 67.7% and 92.3%, respectively. In DLBCL patients, poor performance score at diagnosis, multiple extranodal involvement and involvement of risky areas such as kidney, adrenal gland, muscle, soft tissue, breast, liver, pancreas, paranasal sinus and orbit increase the risk of CNS recurrence during follow-up. In this field, prospective molecular genetic studies focusing on specific pathogenetic mechanisms are needed.